Human motion recognition method and device

Abstract

A human motion recognition method and device are disclosed in embodiments of the present application. A specific embodiment of the method includes: sampling an input video to be identified to obtain asequence of sampled image frames of the video to be identified; using the trained human key point detection model to detect the key points of the sampled image frame sequence, and obtaining the heatmap of the key point position of each sampled image frame in the sampled image frame sequence is obtained, wherein the heat map of the key point position of the human body is used to characterize theprobability characteristics of the preset key point position of the human body; inputting the thermal map of the key points in the frame sequence of the sampled images into the trained motion classification model for classification, and obtaining the human motion corresponding to the video to be recognized. The embodiment realizes the motion recognition by utilizing the coordination and cooperation relationship of the key points of the human body in the video to be recognized and the time continuous characteristic of the human body action, which is favorable for improving the recognition accuracy.

Description

technical field [0001] The embodiments of the present application relate to the field of computer technology, specifically to the field of image processing technology, and in particular to a method and device for human motion recognition. Background technique [0002] Image processing reflects the machine's understanding of space objects and is an important technology in artificial intelligence technology. At present, the processing methods based on static images have made great progress in tasks such as image classification and target tracking. However, for the processing of videos formed by continuous image frames, in addition to the information of the spatial dimension, the information of the temporal dimension is also involved. , the processing complexity is high. [0003] Image processing can be specifically applied in the understanding of human action behavior. In the human action recognition technology based on image processing technology, the human action images ar...

AI Technical Summary

Problems solved by technology

Usually, human actions are constrained by the coordination relationship between body structures and are continuous in time. Each action has contextual information in the time dimension, and it is difficult for action recognition technology based on static image classification and recognition to make use of this information.

Images